To improve the QoS ability of IP networks, Multi-Protocol Label Switching (MPLS) has been created to give QoS guaranteed services. Lately a variety of QoS routing protocols for establishing traffic engineering paths in MPLS networks (IP networks with MPLS) have been suggested and they can be organized into 2 groups: the distributed and centralized approach. In the distributed strategy, a source decides the routing path from itself to its destination and all routing information is kept up to date by flooding; while in the centralized approach, the routing decision is made by the central manager. To find the best routing path, most of the existing routing protocols use the Widest Shortest Path (WSP), which is the shortest path seeking with the thought on hop count and route bandwidth. However, the WSP has high path searching complexity as well as a large dynamic routing table…..
To deal with these disadvantages, we create a new path selection algorithm referred to as Largest Widest Shortest Path with Limited Choices (LWSP-LC) for QoS routing protocols in MPLS networks. The LWSP-LC comes from the WSP though with 2 crucial alterations. To begin with, the number of alternatives to select the optimal path is substantially decreased. Next, an additional parameter, which is the obtainable bandwidth, is considered in the path selection. The LWSP-LC can be implemented in both distributed and centralized approach, and we call the routing protocols as Efficient Distributed QoS Routing (EDQR) and Efficient Centralized QoS Routing (ECQR) respectively. To evaluate the efficiency of EDQR and ECQR with other existing QoS routing protocols, we build a simulation model with 3 different networks including a real network and two traffic scenarios. Our simulation results reveal that, under different situations, our algorithms always have a lower path searching complexity and smaller communication overhead but without significant performance degradation. We discover that the difference of the path searching complexity and the communication overhead could be up to 251 and 46 times respectively. Furthermore, our algorithms do not demand any extra hardware or processing power but the contribution is still significant. Additionally we look into the effect of the range of alternatives on the performance of our routing protocol and simulation results implies that a very small number of choices is adequate to keep an acceptable routing performance in terms of the connection blocking probability. We also build a general rule to compute the upper bound of the minimum number of alternatives for our algorithm in accordance with the network size. By doing this, we can readily set the number of choices for our algorithm as the upper bound of the minimum number of possible choices by using the general rule…..
Source: City University of Hong Kong